Whenever I draw coastlines for a map, I find myself simply drawing large curves and adding in small squiggles in between, occasionally varying things a little for river deltas or small peninsulas. This gets boring after a little while, especially when I do it on a large scale.

  • What are the processes that cause various shapes in coastlines to form?
  • How can I harness these processes to accurately construct bays, peninsulas, river deltas, etc.?

If anyone knows if Slartibartfast is around to talk about fjords, then I'd love to have him answer.


This is part of a series of questions that tries to break down the process of creating a world from initial creation of the landmass through to erosion, weather patterns, biomes and every other related topics. Please restrict answers to this specific topic rather than branching on into other areas as other subjects will be covered by other questions.

These questions all assume an earth-like spherical world in orbit in the habitable band.

See the other questions in this series here : Creating a realistic world Series

  • 3
    $\begingroup$ Excellent Question, we may need to do a challenge on geography... $\endgroup$ – James Jul 27 '15 at 17:56
  • $\begingroup$ @James I agree. How have we not had one yet? $\endgroup$ – HDE 226868 Jul 27 '15 at 17:57
  • $\begingroup$ No idea but I just added it to the topic selection list. $\endgroup$ – James Jul 27 '15 at 18:00
  • 1
    $\begingroup$ Erosion should not be separate from currents and I see no mention of deposition in any of the answers to the second question. Deposition goes hand in hand with erosion and affects coastlines, rivers, lakes, deserts, etc. Tectonics can be safely taken to one side but everything else is a complex feedback mechanism: the shape of the land and sea will affect currents and climate and these will change the shape of the land. And what about the moon? Are the tides to be the same as on Earth? $\endgroup$ – Avon Jul 27 '15 at 18:40
  • 1
    $\begingroup$ This would've been great in the map making challenge $\endgroup$ – Aify Jul 27 '15 at 22:50

It's pretty much random.

But that's a good thing, because it allows for creativity.

The main driving factor here is erosion. The coastal details are random, but some major features depend on the type of land at the coast.

  • First we can discuss a collection of major features for a coast and a bit about how they are formed.
  • Then a couple notes on how to get the big features right.
  • Finally some sources for generating your coastlines automagically.

The features:

enter image description here$_{Source}$

Roughly from left to right:

  • Beaches are formed primarily of the sediment deposited along a coast from the erosion of a surrounding cliff, or bluff.
  • A rocky cliff is made up of material which is relatively resistant to erosion such as sandstone, limestone or granite, a flat rocky wave-cut platform or abrasion platform is formed in front of the cliff.
  • The headland is the land on top of the cliff.
  • (Not listed) A bay is a beach surrounded by headlands.
  • Sea caves are made when certain rock beds are more susceptible to erosion than the surrounding rock beds because of different areas of weakness. These areas are eroded at a faster pace creating a hole or crevasse that, through time, by means of wave action and erosion, becomes a cave.
  • Natural arches are formed when a sea cave is eroded through by waves.
  • Stacks are formed when a headland is eroded away by wave and wind action or a natural arch collapses.
  • Stumps are shortened sea stacks that have been eroded away or fallen because of instability.
  • A skerry is an eroded stack.
  • Spits are formed by ocean currents colliding and depositing materials.
  • A tombolo is formed when a spit connects to an existing island or builds up enough to create an island.
  • A rocky islet is a small island. Perhaps a disconnected tombolo or a isolated skerry.
  • A river estuary is a bay or inlet with a river flowing into it.
  • Sand islands are disconnected spits or the result of river deposits.
  • Dune is a planet with giant worms and spice. Also a pile of sand.
  • A lagoon is separated body of water, usually salty, and occurs quite frequently. Most of these features can occur anywhere, but have a better chance of sticking around depending on what the land is composed of.

Getting some clues:

There are some features/properties that will aid in the selection of features. These aren't rules, but guidelines. For instance:

Reefs and offshore features
An offshore reef or island will significantly reduce the wave strength and you'll get some fairly lackluster beaches. They'll mostly be sediment build up and will likely result in shallow beaches (look at beaches on the Australian mainland inside of the great barrier reef for good examples).

Winds and hemispheres
If there are high waves there will be more erosion and the coastline will more likely have been pushed landward until it hit some rocky materials to slow the encroachment. The high waves are caused by high winds. So, these rocky coasts are characteristic of the west coasts in the northern hemisphere and the east coasts in the southern hemisphere.

If the coastline is filling an area forged by a recent ice age or glacial activity, then large cliffs may be more prolific. The glaciers will have carved out deep valleys, leaving bedrock as the headland which resists rapid erosion. See the fjords in near-arctic regions for examples.

Automagical generation:

Coastlines can be passibly simulated using fractal landscape generation.

I found this article, which describes how the author algorithmically produces a coastline fractal by adding a random component: enter image description here

Also this one doing a similar process:
enter image description here

  • 1
    $\begingroup$ I would only add a mention that tectonics, glaciation, and volcanic activity preceded erosion in this process (basically they form the substrate on which erosion takes effect) $\endgroup$ – James Jul 27 '15 at 19:27
  • 17
    $\begingroup$ +1 for the dune definition, unexpected in the middle of all that science. $\endgroup$ – Josh Jul 27 '15 at 19:45
  • 2
    $\begingroup$ There's a good Ted Talk with Mandelbrot, who is the man whose work the article you linked to is based on. While the coastline may be random, it's randomness, or roughness may actually be measurable. $\endgroup$ – Web Head Jul 27 '15 at 20:17

Speaking from experience:

I began a similar sounding project a few years ago, with layers - arrays of varying sizes, each square, each functioning as its own 2d fluid simulator, labeled 'innercore', 'outercore','mantle', 'crust', 'hydrosphere', and then 'atmosphere'. They mapped values, provided particular methods to describe behavior through the simulation, and impacted on other layers in various ways. I would convert my data into .bmp images for easy testing and then project the data-points as variations of magnitude on simple spherical-shaped objects on an OpenGL application. On the crust layer, the most detailed, I'd have multiple datapoints per 'pixel', for elevation, change in elevation, granularity(for determining erosion rates and textures), and some other values for path-finding and logic (basically bit-crunched surface normals).

Using this method, I was able to generate simple and convincing landmasses by pausing at various times during the simulation. I wasn't satisfied with static maps because I was modeling populations and genetics and I thought it was silly the map would be static and unmoving so I ditched the Civ-style fractal generators and, imho, was able to generate more convincing terrain features with the simulator. My atmosphere layer reacted to elevation of the crust and 'evaporation' off exposed sections of the hydrosphere, thus producing rainfall patterns and watershed formations. My hydrosphere gave me oceans, tides, and a water-table which in turn defined the rivers and lakes. Because they all used roughly the same methods, akin to fluid dynamics, I didn't have to pull my hair out being explicit about defining and sub-classing things such as Mountain_Chain, River_Delta and quereying Pathfinding.riverpath and other such nonsense. It all occurred, for lack of a better term, naturally, arising from the simulation. Erosion was achieved through a 'granularity index' (think solid-rock, boulders, slew, sand, etc) and this caused canyons and river valleys to form, and even gave me things like sandbars and dunes. I was working on a method for interpolating values (cause I wanted to zoom) when I got distracted by a rare encounter with a female. Eventually, when I returned, my OCD began to kick in and I never completed the methods for the interpolation of 'fine' details because I became increasingly and inconsolably perturbed by subtle frustrations resulting inevitably from attempts at mapping a SQUARE set of data over a CURVED surface. I tried everything, and everything failed me. I once, caught up by the vanity of youth, even became a vocal enthusiast for something called a quincunx. What a fool i've been!

Eventually I found this and decided I should re-write the simulator using discrete points along an Archimedean spiral. I sorta burnt out on that project, though. it showed much promise! I recommend - highly - to do away with those tired old fractal-based generators - this isn't the 90's anymore! EVERYTHING is a fluid at various states of arrest - so to me it makes sense to use a fluid simulator. You should map the data using the methods they describe cause it's OCD approved. And singularity-proof. Although, you'll have to learn how to invert incomplete elliptic integrals of the second kind, I'm sure you'll figure it out. Ain't no hill for a stepper!

  • 3
    $\begingroup$ ugg, I sound like a maniac. $\endgroup$ – punkerplunk Jul 27 '15 at 22:07
  • $\begingroup$ Just a note on those interpolation functions - instead of using random data, like the fractal geeks, the idea I was leading toward 'reflected' the larger geometry upon the smaller scales. Coastlines, formed from the intersection of elevated portions of crust and the hydrosphere at sea level - was defined completely by topography - a line of intersection between two planes. $\endgroup$ – punkerplunk Jul 27 '15 at 22:30
  • $\begingroup$ Sounds like a fine method for simulating minor changes over a (lacking plate tectonics) short time. But I don't see where you start from. Don't you need to draw a map with all geographic features (at minimum the elevation of all pixels and water table) before you can begin this simulation? $\endgroup$ – Samuel Jul 27 '15 at 23:48
  • 3
    $\begingroup$ @punkerplunk Please post some screen shots of this. I'd love to see it. It honestly sounds more like a wishlist of features rather than something a single person actually implemented in their spare time. $\endgroup$ – Samuel Jul 28 '15 at 2:29
  • 1
    $\begingroup$ @punkerplunk, I'm a professional .NET (amongst other things) programmer. If I was to do anything like this, it'd be in my spare time, probably an hour or two a day at most. This might make a good community project, I'm sure there were some other programmers here. $\endgroup$ – Monty Wild Jul 28 '15 at 3:18

If this is the proposed method:

  1. Do a rough global land mass.
  2. Divide it into plates.
  3. Move them around - creating mountains where the converge on land and chasms where they diverge and sharp breaks where they slide past each other.
  4. Decide that these bits will erode from glaciers and these bits will erode from coastal flows (and these bits by rivers?)
  5. Decide on a global climate, ocean currents and rainfall and rivers (not decided previously) and therefore vegetation etc.

then I think that is both a hard way to do it and likely to produce poor results without an enormous amount of effort. Steps 3, 4 and 5 should all be happening at the same time. Tectonic movements can and do make significant changes on the timescale of erosion and deposition. The changing shape of the land will affect coastal flows, rivers and weather systems. I don't think you will be able to adequately simulate the intricate processes of geology like that.

I suggest a slightly different approach:

  1. Do a much less rough map. Ideally generate it using the techniques of fractal landscapes (there are very sophisticated tools available for this), which will produce much more realistic randomness than a human could:

A fractal map from civfanatics.com

  1. (Optional) Copy (imprecisely - the less precisely the better) some real features from real maps (from Earth or elsewhere). They can be distorted and in sections. Put them wherever you like; wherever looks right.

  2. Decide on plate boundaries that will have lead to that situation. E.g. mountain range here implies a convergent boundary; chasm here might be a good place for a divergent boundary; sharp edge here, maybe a transport boundary.

  3. Tweak the map in accordance with those boundaries: make mountain ranges on convergent boundaries more ranging. Between two convergent boundaries there will be a divergent boundary. Put volcanoes and chasms there if they weren't already. Volcanic islands if it's out to sea. And so on. Remember plates can hinge/rotate. A convergent boundary of a rotating plate will produce a mountain range of increasing size. It's far side will be a curved transport boundary.

  4. (Optional) Gently tilt, elevate, and depress some areas of land (whole continents) along plate lines - submerging into the ocean or raising the sea bed on to dry land.

  5. (Optional) Add the results of ice ages (gouged out valleys near the poles) and rocky deposits at their extremities.

  6. Decide on sensible global winds and ocean currents and temperatures.

  7. Smooth out rough bits that protrude into those currents and add smooth deposition places further down stream of those currents - where the current's course changes, slows or becomes shallower.

  8. Decide on rainfall based on the air currents, elevations and temperatures and make sure rivers lead from them downhill to the sea with lakes along the way if they pass through depressions.

  9. Do some erosion and deposition (flood plains, deltas, etc.) for the rivers.

  10. Decide on vegetation based on climate, rainfall and proximity to rivers, lakes and coasts.

The more care you put in the better the results will be but I think good results will come from a lot less effort then the 'from scratch' approach.

  • $\begingroup$ This is okay, but the question states, "Please restrict answers to this specific topic rather than branching on into other areas as the other stages will be covered by further questions." I know that many factors are interrelated, but not all are. $\endgroup$ – HDE 226868 Jul 27 '15 at 19:43
  • $\begingroup$ @HDE226868 Indeed. I'm saying the entire approach is wrong for 'Creating a Realistic World Map'. And that's especially true for the existing three questions and answers: erosion is answered before currents, no mention of deposition. Reality is the result of all things happening at the same time. Volcanic islands appear on the same time-scale or faster than erosion and deposition for example. Trying to simulate it to produce a world is a non-starter in my opinion. $\endgroup$ – Avon Jul 27 '15 at 19:53
  • $\begingroup$ The steps are not being done in any particular order. The questions aren't simultaneously asked, but the processes may still be. $\endgroup$ – HDE 226868 Jul 27 '15 at 19:59
  • $\begingroup$ @HDE226868 But still I think the better approach is to start out with a realistic fractal map and then explain it with those processes and tweak it accordingly. Rather than try to generate a map with those processes. $\endgroup$ – Avon Jul 27 '15 at 20:01
  • $\begingroup$ That's fine. I might use the fractal approach in the future. $\endgroup$ – HDE 226868 Jul 27 '15 at 20:02

Could this be like something you're after?

Computer generated terrain with plate tectonic simulation

For just coastlines, I think fractal methods are better, but if you want the entire topography, I recommend plate tectonics, e.g. as in the PlaTec project.

  • $\begingroup$ Regrettably no one reads the idea at the bottom of topic... $\endgroup$ – Shadow1024 Oct 30 '16 at 21:23

I can offer a quick and dirty method that relies on the fact that coastlines and other geographical features are fractal in nature and therefore show self-simularity in all scales i.e. a coastline looks the same at the scale of one meter as it does at one kilometer.

Unless the human eye has something other than the fractal shapes to provide scale e.g. a vehicle, city, roadway etc, geographical features on images look realistic even after you scale the image up or down them up or down. The higher the resolution image you start with, the more you can scale up while keeping the realistic fractal properties.

So nip over to google maps/earth of some other similar services and start looking for islands. I suggest Indonesia, the Philippines, the Ageian, the coast of Alaska and B.C. etc. Gobs of islands in all shapes and sizes, most of which no one who isn't from there would recognize.

Pick some small obscure island, grab as high resolution image of the whole island as you can, image, then load it into pop into your favorite image editing app. To obscure any possible familiar outline, rotate the image to some arbitrary angle other than the one holds in reality and/or mirror it right-to-left or up and down. Somewhat surprisingly, I've found that where so used to seeing land masses in specific orientation that even a minor change in orientation makes the unrecognizable to most.

Then scale the image upward until it reaches the scale on you map of the landmass you want and presto, instant novel land mass continent with realistic coastlines and usually all other terrain features as well.

The same basic trick works on all geographical features because they all have fractal properties just like coast lines. Creeks scale to the Amazon and often hills make passible mountains (although to always, some hills are degraded mountains but others result from other actions like glaciers. In many cases you might find the to rounded.) If they don't you can always swipe mountains of craggy hills from some other part of the globe, rotate, mirror and scale and drop it on the continent.

You can also with some practice, create a union of two resized islands together to make a completely novel profile although, making it look seamless is tricker than than it sounds. In this case, the fractals are working against you. It's easy to get plains and other flatlands to merge fairly transparently but usually not mountains or other rough terrain because the transition in scale becomes apparent along the seam.

Also, don't forget coastal plates under the water. The edges of the plates look very much like coastlines for the undergo a very similar process of erosion. They even have rivers and channels.

In thinking about it, I've only done this in 2D but will the 3D mappings out there, you might be able to pull the same trick in 3D to make a fly over.

It's been a while since I used this method to cobble together a quick world for a friend so I apologize for not having a handy example at hand but I think the idea simple enough for everyone to grasp it.

At the very least, it's a good way to quickly prototype a world in preparation for for using a more sophisticated method.

Gotta love fractals.


A quick way I have found useful is to look at an actual map and take country, states, lakes, and other real life formation and invert them, flip them, rotate them etc... Ex: Trace the Black Sea outline then rotate it 90 degrees. You now have a kidney shaped continent. The "water" would be your actual land mass and the surrounding land would become your oceans.

Many famous authors do this.

Westeros (Game of Thrones) looks very much like England. Its just larger(about the size of South America)

Warhammer (Games Workshop) is extremely close replica of our actual world with slight changes.

Another series ( i forget the name of it) basically used Alaska and inverted it.

Also the leaf method. Take a leaf(preferably a large one). Crumble it a bit on the edges to get the jagged appearance. This is easier in the Autumn. Then trace that edge pattern rotating around the map.

There is a cool Youtube channel (Questing Beast) that does a lot with map drawing. Here is an idea for dropping sugar cubes and tracing the pattern. (Unique way. he has great mountain and forest stuff as well.)


After you get the coastal boundaries then you can begin to add everything else.


First, consider the steepness of the land off the coast.

How quickly does the land drop off into the ocean? The faster it drops off, the more likely you are to have steep cliffs along the coast, maybe a few sea stacks, and then open ocean. Long archipelagos trailing off from the coast will be found in regions where tectonic boundaries run perpendicularly away from the coast line.In places where the sea slopes off over a longer distance to the edge of the continental shelf, the coast will have more large-scale structures. Alaska, for example, has a mountainous coast with lots of islands, but is close to the edge of the Pacific plate, and hence runs in a fairly straight line on a larger scale, with no major protrusions like the Canadian Arctic Archipelago or Florida. In addition, structures like barrier islands are generally more common where the land slopes gently into the sea.

Next, look at the general land forms just inland from the ocean.

On a slightly smaller scale, one of the biggest drivers of coast line structure is the type of land that it forms next to. The coast of Southern Alaska, for example, lies next to a large range of mountains. This geological unevenness creates a fairly broken looking coastline with lots of large islands along the coast. On the other extreme, the coastlines of southern India slope gently into the sea, forming few islands and lots of long, smooth sandy beaches.

The coastal features found in a particular region will also depend heavily on the climate type. Flatter regions will form flat, sandy spits and barrier islands, while steeper coasts will form rocky outcroppings and mountainous offshore islands.

Look at erosive and constructive processes that will modify your coast.

Volcanoes, glaciers, and rivers can all change a coast line. Lava pouring into the ocean can extend and smooth out coastal features, filling in small valleys and pouring into bays, while at the same time creating longer outcroppings of rock in the places where lava enters the ocean. Glaciers can form long, steep valleys cutting into the coastline, known as fjords, and rivers can form either long inlets cutting into a coastline or build deltas extending out into the ocean, with the former occurring more often in deep waters and the later forming on top of shallow continental shelves.

Following these steps should give you a coast line that looks more 'real', and will make maps more varied than applying random noise equally to every region of your map. You can, of course, go into more detail about how a coast forms, with areas with lots of sand accumulation forming dunes, and different erosion patterns dominating a section of the coast based on the orientation of underlying rock strata.


I checked briefly the other questions and my answer won't be anywhere near the same caliber (I think) - but to get things going:

Drainage basins and elevation

I normally always start with drainage basins and their rivers and lakes, and base both my topography (well those two are a bit more simultaneous than one before the other in my case) and coastlines off it. That's what determines my bays and deltas and fjords etc...

Afterwords it goes with the coastal elevation, higher elevations close to the sea will have a more jagged coastline, steeper slopes have tiny streams rivers dumping into the sea so those also act in forming it, where lower terrain will be 'smoother'?

Obviously the scale of the map is going to change a huge amount of things.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.